**** BEGIN LOGGING AT Tue Nov 09 02:59:58 2010 Nov 09 03:00:22 seems easier to just use eglibc for my distro instead -- i'm using angstrom 2008 Nov 09 03:07:50 mrj10, you should just use eglibc unless you have a good reason to need glibc instead Nov 09 03:09:11 i agree, i just went with the default for angstrom 2008 for simplicity's sake, i overrode it in my local.conf and am rebuilding now Nov 09 03:31:00 trying to compile eglibc, i get the same error as this guy: http://permalink.gmane.org/gmane.linux.distributions.gumstix.general/53998 Nov 09 03:33:34 mrj10, i suppose there's a reason angstrom.2008 is still using glibc. Have you tried angstrom.2010? Nov 09 03:34:30 i have before. i reverted back to try to fix something, but i found the cause to be something else, so i may switch back to 2010. yarg. Nov 09 03:35:01 maybe i can get away with just switching the gcc version to the one used in angstrom 2010 Nov 09 03:35:21 heh. angstrom.2008 uses older versions of toolchain bits, so they have probably had much less testing Nov 09 03:35:36 (with eglibc, that is) Nov 09 03:35:41 yeah Nov 09 03:36:06 i'll try ANGSTROM_GCC_VERSION = "4.5" in my local.conf. if that doesn't work, i'll nuke the whole thing and go angstrom 2010 Nov 09 03:36:27 angstrom.2010 has eglibc 2.11 Nov 09 03:36:40 you might find that works better. Nov 09 03:36:47 well, true, i guess i'll do that first Nov 09 03:37:04 and, if you are inclined to bump gcc, you may need to bump binutils as well... Nov 09 03:37:20 * grg shrugs Nov 09 03:38:27 thanks for the help, one of those or the nuclear solution should work Nov 09 03:38:39 no probs Nov 09 04:03:54 grg: eglibc 2.11 appears to compile with angstrom 2008's gcc :) Nov 09 04:04:54 nice Nov 09 06:03:48 i've been mucking around a bit with an image .bb file and a task .bb file that includes most of the packages needed for the image Nov 09 06:04:33 somewhere in the process, opkg has lost its head and doesn't what packages are around, so when i do the do_rootfs() task on my image recipe, it craps itself and says it can't satisfy a bunch of dependencies even though the .ipks are all sitting there Nov 09 06:04:46 anyone know how to get it to rebuild its knowledge of the world? Nov 09 06:14:16 mrj10: try bitbake package-index Nov 09 06:14:51 khem: will do Nov 09 06:26:10 khem: no love :( Nov 09 06:27:19 ah, i see Nov 09 06:27:27 hmm Nov 09 06:28:02 if i switch angstrom's libc from glibc to eglibc in my local.conf, do i need to rebuild every package? Nov 09 06:28:40 or can i do some kind of symlink atrocity and symlink all the packages in deploy/glibc/ipk into deploy/eglibc/ipk? Nov 09 08:56:09 elo Nov 09 08:57:41 hi hrw Nov 09 09:01:21 * XorA growls Nov 09 10:47:26 hi, is there any work in progress in OE on using linaro's gcc 4.5 for a complete hardfp system? Nov 09 10:47:48 Angstrom had hardfp before Linaro existed Nov 09 10:48:29 XorA: using non-lite codesourcery toolchain? Nov 09 10:48:34 XorA: sorry again for the wrong link. I was lazy and copied RP's last proposal without thinking Nov 09 10:52:06 I think we used upstream gcc Nov 09 10:52:17 not sure anyone ever booted the result Nov 09 10:52:34 but it should be possible to repeat if you look in the angstrom config files Nov 09 10:53:13 stefan_schmidt: no probs, just thought I would try and shoot it down in a memorable way so people remember Nov 09 10:53:30 stefan_schmidt: the issue with pointing at KDE is at any random point KDE could vote to change the rules Nov 09 10:53:33 XorA: yeah, best approach :) Nov 09 10:54:25 I shoudl have surrounded it in tags :-D Nov 09 10:55:27 ;) Nov 09 10:55:56 orjanf: for the current patche we use from linaro you could also peak in the gcc 4.5 recipe Nov 09 10:57:16 if gcc 4.5 is already pactched then it should be real easy to do an Angstrom build with hardfp enabled Nov 09 10:59:17 XorA: I haven't tracked things closely enough to say if their compiler patches requires (e)glibc patches and/or kernel patches (in case the userland/kernel ABI changed wrt syscall parameter passing or similar) Nov 09 11:00:05 for glibc I've only seen their WIP cortex-strings stuff which of course could be integrated, used separately, or not at all Nov 09 11:05:07 orjanf: I think you just volunteered :-D Nov 09 11:06:14 XorA: maybe, time permitting :) Nov 09 12:21:22 khem: http://wiki.openembedded.org/index.php/Donations Nov 09 12:25:24 I'm trying to use all files in my reciept's subdirectory /files/ as the source files to compile. SRC_URI = "file://${FILE_DIRNAME}/files/" but bitbake shows me an error in do_distribute_sources stage. Is there any way to make such a simple thing without packing source files? Nov 09 12:28:25 dv: see srctree bbclass Nov 09 12:34:17 Good evening Nov 09 12:36:55 I baked the task-sdk-host, it gives me binutils-cross-sdk ipk for installing on the host Nov 09 12:38:27 But I also need the library files as in "sysroots/armv6-angstrom-linux-gnueabi/usr/lib". Is there a recipes for packaging these? Nov 09 12:51:11 heh Nov 09 12:51:40 apparently we also need to go through the statutes and remove references to kde :) Nov 09 12:56:34 on angstrom 2010 (gcc 4.5), i get errors compiling ti-dsplink http://pastebin.com/d8E2BatT (undefined reference to '__aebi_uidivmod' ) Nov 09 12:57:11 i seem to remember getting this error before, i switched back to angstrom 2008 to get a different toolchain version as a bandaid fix, but curious how to fix it in a more satisfying way Nov 09 13:11:52 ah, it is this: http://comments.gmane.org/gmane.linux.distributions.angstrom.devel/4503 Nov 09 13:12:19 how can i override the BINUTILS version just for that package? is it possible? Nov 09 14:07:04 khem: Hi ... did you fix the eglibc build issue? Nov 09 14:36:28 for those interested, the TI compilation issue with binutils 2.19+ seems to be the following: http://www.eclipse.org/forums/index.php?t=tree&th=197704&S=bf2c36fb547f9973f300fdf573c34d81#page_top Nov 09 14:45:46 hmm, are we going to end up moving to gitolite from gitosis at some point, i wonder? Nov 09 14:51:11 kergoth: first I've look at gitolite -- seems pretty nice, and would give us some features we need like rewind protection Nov 09 14:52:14 yeah, seems quite feature rich. i really like the access control within a repository, conceptually, though I don't know how much we'd want to leverage it Nov 09 14:53:14 * kergoth ponders Nov 09 14:55:59 cbrake: is there a way to hide a repo on our git server from the web interface, via gitosis-admin, do you know? thinking for testing / prototype stuff Nov 09 14:56:38 kergoth: yes, you have to explicitly export repos with cgit via a config file Nov 09 14:56:47 ah, okay, cool Nov 09 14:57:22 I was thinking about putting a source/m4.git up there for experimentation with a recipe that just clones already patched sources from a git repo within our control, rather than relying on upstream directly Nov 09 15:15:29 Have you guys looked at gerrit at all? It does repo hosting, branch level access controls, and code reviews. Nov 09 15:16:03 I'm setting it up now for use within our company - seems pretty good. Nov 09 15:17:58 hm so you only need to deploy the war into a glassfish Nov 09 15:18:58 I know nothing about java stuff. Gerrit comes with an inbuilt daemon with ssh and http server. You can proxy it/whatever if you want/need. Nov 09 15:19:08 hu? Nov 09 15:19:26 ssh in a war? Nov 09 15:19:59 * Jay7 fear it's resource usage.. Nov 09 15:20:10 jay7 hm Nov 09 15:20:34 +1 for providing postgres Nov 09 15:20:38 binding Nov 09 15:21:53 * kergoth rather likes reviewboard more than gerrit for code reviews, but hasn't used either that much Nov 09 15:21:58 well.. time to break things Nov 09 15:25:52 * ka6sox shudders Nov 09 15:29:19 * mwester wonders if we should just switch to code.google.com Nov 09 15:29:30 :p Nov 09 15:33:10 * kergoth rolls eyes Nov 09 15:33:52 hi ka6sox Nov 09 15:34:05 ka6sox whats the problem with j2ee? Nov 09 15:36:03 * ka6sox dunnos Nov 09 15:36:47 I was commenting on Jay7's comment about breaking things. Nov 09 15:37:17 hehe :) Nov 09 15:37:18 * kergoth mulls over different workflows for managing packaging in a git repo based on various links on vcs-pkg.org Nov 09 15:37:28 * Jay7 is going to resize lvm partitions Nov 09 15:38:04 that reminds me, is there a nice easy tool for that? resizing the filesystems and partitions at the commandline by hand is annoying Nov 09 15:38:21 kergoth: I'm going to do it by hands.. Nov 09 15:38:42 did you tried gparted? Nov 09 15:38:59 btw.. Nov 09 15:39:04 let's try gparted :) Nov 09 15:40:01 not yet :) Nov 09 15:40:04 good luck :) Nov 09 15:40:18 I'll try then ;)) Nov 09 15:40:21 kergoth gparted Nov 09 15:40:26 cool Nov 09 15:40:56 btw Nov 09 15:41:13 is anyone have used ext4 here under OE TMPDIR? Nov 09 15:41:44 I did Nov 09 15:41:50 never saw problems Nov 09 15:41:58 I'm using reiserfs now Nov 09 15:41:59 how about speed? Nov 09 15:42:07 considering about switching to ext4.. Nov 09 15:42:25 ka6sox: I run builds in fire/forget way Nov 09 15:42:26 bbl Nov 09 15:47:29 jay7 reiserfs? Nov 09 15:47:32 oh my god Nov 09 15:48:34 woglinde: what? :) Nov 09 15:48:54 it is very fast and stable :) Nov 09 15:49:00 but very old too.. Nov 09 15:49:52 for me reiserfs was not stable Nov 09 15:51:45 gparted doesn't support lvm atm :( Nov 09 15:52:05 0.7.0 at least Nov 09 15:56:44 Hi Nov 09 15:57:11 Does bitbake builds new configure before compiling? Nov 09 15:57:49 Right now my package only works if manually run buildconf --force before doing do_configure Nov 09 15:58:09 the autotools class runs autoreconf before running configure Nov 09 15:59:03 The problem, patch is applied to acinclude.m4 Nov 09 15:59:21 But configure script stays the same, meaning new configure was never generated. Nov 09 15:59:42 I do have: inherit autotools Nov 09 15:59:59 patches to acinclude.m4 are wrong Nov 09 16:00:20 Why is that? Nov 09 16:00:49 because its generated Nov 09 16:01:10 putting all m4 macros into one file is bad anyway Nov 09 16:02:08 I thought that acinclude.m4 is used for generating configure (also bunch of other files). Nov 09 16:02:27 So its best to do patch on acinclude Nov 09 16:03:31 hm ah sorry aclocal Nov 09 16:03:36 was the included filed Nov 09 16:03:38 ups file Nov 09 16:03:54 but autoreconf should generate the new configure Nov 09 16:04:44 03Tom Rini  07org.openembedded.dev * rae63b85341 10openembedded.git/recipes/gzip/gzip_1.4.bb: Nov 09 16:04:44 gzip: Fix postrm section, bump PR Nov 09 16:04:44 We don't want to un-rename files here since they're packaged up Nov 09 16:04:44 as renamed Nov 09 16:04:44 Signed-off-by: Tom Rini Nov 09 16:05:11 Give me a sec, have to make few tests. Nov 09 16:07:29 woglinde, here: Nov 09 16:07:45 I run unpack, patch and configure. Nov 09 16:08:10 Checking configure md5sum after each of these steps. Nov 09 16:08:13 Its the same. Nov 09 16:08:23 New configure file is not generated. Nov 09 16:08:53 Wait.. Nov 09 16:09:11 I have do_configure_append() and I have do_configure() [calls oe_runconf] Nov 09 16:09:25 Maybe these make some problem? Nov 09 16:09:40 please pastebin your recipe Nov 09 16:11:15 http://pastebin.com/hwMBay7N Nov 09 16:12:30 try to remove the do_configure Nov 09 16:12:47 davidlt: you're overriding autotools.bbclass. Nov 09 16:12:55 your recipe is explicitly only running ./configure Nov 09 16:13:00 so yes, it wont be regenerating configure Nov 09 16:14:19 http://docs.openembedded.org/usermanual/html/chapter_reference.html#autotools_class Nov 09 16:14:32 autotools_do_configure is never called. Nov 09 16:14:52 I get it now. Thanks for help! Nov 09 16:14:55 Can I pull all the files needed to bitbake a toolchain into one place ? Nov 09 16:15:42 atl: what do you mean? Nov 09 16:15:59 I think he means do_fetch Nov 09 16:16:02 for all packages Nov 09 16:16:20 I would like to archive my build, yes would fetch be enough? Nov 09 16:16:30 that depends on what you mean by "archive my build" Nov 09 16:16:38 what exactly are you trying to accomplish? fully offline build? Nov 09 16:16:49 pull network cable and still be able to bitbake meta-toolchain Nov 09 16:17:20 yes, fully offline build would describe it Nov 09 16:17:48 then yes, fetchall is all you need, that will fetch everything from upstream into your DL_DIR Nov 09 16:19:20 if I have done a build and have a populated DL_DIR, then I wouldn't do a compile next and manually go thru the commands? Nov 09 16:19:25 guh, there are too many issues with sharing source trees between recipe variants. even when using separate objdir there's a problem, if the recipe does things like sed files in ${S} rather than ${B} in tasks other than unpack/patch Nov 09 16:19:34 atl: what do you mean? Nov 09 16:19:38 oops s/wouldn't/would Nov 09 16:20:08 oe never downloads files its already fully downloaded, whether you're starting from an empty tmpdir or not Nov 09 16:20:21 once DL_DIR is populated, you can build away happily without it ever going online Nov 09 16:21:39 unboundedlocalError: local variable 'cooker' referenced before assignment Nov 09 16:22:12 thanks, knowing you think this is possible I'll try and figure that error out. Never done this before, thank you Nov 09 16:28:03 atl: if you can pastebin that cooker error, i can investigate Nov 09 16:28:15 atl: would need to know what bitbake version you're on, though Nov 09 16:29:24 I don't think its a real issue, bad user on device error, will clean things up and try again. Nov 09 16:30:19 no, that error indicates a bug Nov 09 16:30:29 perhaps you can find a way to not enter that codepath, but its still a bug Nov 09 16:31:16 hi gnutoo Nov 09 16:32:12 khem: ping Nov 09 16:33:18 hmmm Nov 09 16:33:36 I'm building one image based on console-image. In the first build, everything works! When I try rebuild with some changes I can't init my system. udev isn't loaded and after kernel boot my init is never executed Nov 09 16:33:42 its running now, I am using angstrom and didn't remove the build/conf/local.conf so bitbake was lost in the wrong directories. Nov 09 16:35:32 hrm Nov 09 16:36:46 hm qt 4.7.1 is out Nov 09 16:37:08 wtf, tarfile.ReadError in package_update_index_ipk.. anyone seen this one? Nov 09 16:37:15 no Nov 09 16:38:05 * kergoth wonders if oe recipes should be in band or out of band for a git src repo, no patch set Nov 09 16:38:07 * kergoth ponders Nov 09 16:42:19 Anyone that's on the members list around? Nov 09 16:44:07 kergoth: I've found 2 GUI's to lvm: kpvm and system-config-lvm Nov 09 16:44:14 first is kde4-based Nov 09 16:44:24 looking on it now Nov 09 17:19:26 anyone looked at the problem sakoman reported on the list Nov 09 17:36:29 Crofton: I think folks just automatically ignore anything from me :-) Nov 09 17:37:43 heh Nov 09 17:37:48 so many emails Nov 09 17:37:51 so little time Nov 09 17:37:55 I will have to fix it Nov 09 17:38:00 is something up with the openembedded-devel mailing list? Nov 09 17:38:20 haven't gotten anything from there in 3 hours, that never happens! also, i sent mail to the list a couple hours ago Nov 09 17:38:43 I think something is up with the list serve Nov 09 17:38:50 I'm waiting on a post I sent to oe-members Nov 09 17:40:43 hm ka6sox? Nov 09 18:06:29 florian, ping Nov 09 18:06:36 hi crofton Nov 09 18:06:56 Crofton: pong Nov 09 18:07:08 hi florian Nov 09 18:07:26 florian, I think the oe-members list is clogged up Nov 09 18:07:40 If I start to talk bullshit that's because I worked on making Android work on some board the whole day. Nov 09 18:07:47 oops Nov 09 18:08:06 ok Nov 09 18:08:12 vent away Nov 09 18:08:57 is see... filesystem full :-/ Nov 09 18:09:02 bummer Nov 09 18:09:08 time to get narcissus moved ;) Nov 09 18:09:12 yeah Nov 09 18:09:19 hm Nov 09 18:09:34 I've been talking with koen and ka6sox Nov 09 18:09:45 sounds like we can move some stuff to osuosl Nov 09 18:10:11 need to look at the issue sakoman reported Nov 09 18:10:39 ok, works again Nov 09 18:19:55 yay Nov 09 18:20:02 florian, thanks Nov 09 18:20:39 unless someone hacks at ifupdown rhat is competent, I am going to try it :) Nov 09 18:21:28 hm no script which is mailing at critical diskspace? Nov 09 18:21:33 cannt believe it Nov 09 18:21:44 time for logwatch? Nov 09 18:23:56 woglinde: feel free... ltg is a community project Nov 09 18:27:04 * Tartarus would recommend logwatch then Nov 09 18:28:15 on debian/related, you need to fix /etc/cron.daily/00logwatch so it mails where you want it to Nov 09 18:28:21 dunno about rhat Nov 09 18:30:57 nagios :) Nov 09 18:31:34 * Jay7 will setup zabbix soon to monitor customers servers Nov 09 18:36:03 hmmm. i wonder how much work it would be to create a fully functional 'native' machine, including merging native/cross/target recipe instances Nov 09 18:39:05 you mean build OE in a rootfs made by OE? Nov 09 18:39:15 native-sdk should provide that, today Nov 09 18:40:23 yes, but i mean without building crap unnecessarily Nov 09 18:40:43 a single 'm4' recipe is all you'd need for native and non-native both, out of the single build Nov 09 18:40:56 ? Nov 09 18:41:34 hm Nov 09 18:42:08 for a true native machine, we don't need oe to build a "crosscompiler", etc. all unnecessary. we'd need glibc, or an 'external toolchain' that extracts from /usr, of course, but beyond that.. Nov 09 18:49:02 Ah Nov 09 18:49:07 I think I see what you're driving at Nov 09 18:58:44 internet connection is worse there days.. :( Nov 09 18:58:56 well.. lvm was resized successfully Nov 09 18:59:08 now I have special partition for TMPDIR Nov 09 18:59:17 and I can clean it by newfs :) Nov 09 19:00:36 BTW, JFYI, find tmp/* -delete is much faster than rm -rf tmp/* Nov 09 19:00:44 weird Nov 09 19:01:29 well.. may be not a very much but faster anyway :) Nov 09 19:02:23 still strange Nov 09 19:05:01 now I should assemble container back and may be describe this setup somewhere on wiki Nov 09 19:16:04 jay7 ugh Nov 09 19:16:28 but I noticed too removing a dir is faster than /* Nov 09 19:17:08 woglinde: our $TMPDIR have about 10 entries Nov 09 19:17:27 so this shouldn't affect Nov 09 19:17:43 difference is somewhere in algorythm Nov 09 19:18:41 and this may (and will) different on different FS, imho Nov 09 19:22:07 hi all Nov 09 19:22:44 kergoth: FWIW when I revert that patch from bitbake it works all fine Nov 09 19:22:51 weird. Nov 09 19:22:52 makes no sense Nov 09 19:23:00 bitbake world has been running all ok since last 24 hours Nov 09 19:23:11 yeah I wonder Nov 09 19:23:27 atc? Nov 09 19:23:43 jay7 hm has soemthing todo with the recursive stuff Nov 09 19:23:48 ARMTechnology Conference Nov 09 19:24:59 intressting Nov 09 19:25:08 is that arm's home fair? Nov 09 19:26:32 lol Nov 09 19:26:34 ARM Powered Corvette Nov 09 19:28:42 its bigger than what I thought Nov 09 19:28:58 hm santa clara thats near your home right? Nov 09 19:32:31 woglinde:yes and thats why I am attending Nov 09 19:32:34 :) Nov 09 19:33:17 already thought that Nov 09 19:35:22 kergoth: I think we could use the patch ericben posted for the graphviz problem Nov 09 19:36:15 jo setfan Nov 09 19:36:19 hi woglinde Nov 09 19:36:21 hi all Nov 09 19:36:37 hms need to root my device Nov 09 19:37:36 woglinde: You bought the wrong one if it is yours and you are not root :) Nov 09 19:39:37 he Nov 09 19:40:07 there is only one device with a9 and tegra Nov 09 19:40:27 woglinde: wetab Nov 09 19:40:35 wetab is atom Nov 09 19:40:47 right but I meant some other successor Nov 09 19:40:53 khem|atc: how is the ARM conference going? Nov 09 19:41:06 stefan_schmidt:its heating up :) today is day 1 Nov 09 19:41:08 by the way when booting ubuntu I am root Nov 09 19:41:21 woglinde: :) Nov 09 19:41:23 khem|atc: good Nov 09 19:41:52 there are few tracks on Open Source Nov 09 19:41:56 03Khem Raj  07master * r6f5c792dd7 10openembedded.git/recipes/gcc/ (16 files): Nov 09 19:41:56 gcc-configure-common.inc: sed /usr/include hardcodes paths to use target_includedir Nov 09 19:41:56 This fixes the problem where gcc fails to detect presence of libssp Nov 09 19:41:56 and does not detect powerpc long double 128 bit support correctly on systems which do Nov 09 19:41:56 not use common /usr/include paths like we have in micro distro Nov 09 19:41:57 Signed-off-by: Khem Raj Nov 09 19:41:58 Acked-by: Tom Rini Nov 09 19:42:05 but not on android Nov 09 19:42:17 I have few linaro folks here from ARM I talked to Nov 09 19:43:03 khem|atc: good, I hope they will do a lot of upstream work for ARM in the future we could rely on in OE :) Nov 09 19:43:59 yeah I told them that we are using the stuff they have and they were quite happy as they think thats what they intended linaro to do Nov 09 19:44:35 re Nov 09 19:44:43 khem|atc: excellent Nov 09 19:44:59 * khem|atc moved the blog website to use wordpress Nov 09 19:45:26 khem|atc: make sure your php and wordpree is up to date all the time when doing this :D Nov 09 19:45:28 hrm Nov 09 19:45:28 * khem|atc is much happier with wordpress thus far Nov 09 19:45:44 woglinde_: OT, are you happy with alice in berlin? Any good local alternatives? Nov 09 19:45:48 I dodnt trust php Nov 09 19:45:54 stefan nope Nov 09 19:45:59 * stefan_schmidt needs to do some research on this when moving to berlin in one year Nov 09 19:46:04 hmm so what do u use CGI/perl Nov 09 19:46:07 or plone Nov 09 19:46:11 grr Nov 09 19:46:20 if your cables is rotten the provider is not important Nov 09 19:46:28 khem|atc: nanoblogger, static html pages generated on my laptop and synced via ssh :) Nov 09 19:46:39 my web-fu is whitebelt and with wordpress I seem to get what I need Nov 09 19:46:40 stefan_schmidt let us collect the gifts Nov 09 19:47:28 khem|atc: is that bug only with -k world? Nov 09 19:47:28 fidencio[AWAY], hi pespin_ improved a lot the recipes Nov 09 19:47:39 khem|atc: I don't have web skills either. But nanoblogger supports markdown and does all the formatting for me. But it is a niche solution i agree Nov 09 19:47:47 woglinde_: heh Nov 09 19:47:56 kergoth: I only can try -k Nov 09 19:48:07 kergoth: because it does not go that far without -k Nov 09 19:48:07 well, i mean more that its specific to world than to -k :) Nov 09 19:48:19 * kergoth nods Nov 09 19:48:32 kergoth: I think its not related to world Nov 09 19:48:42 someone else saw it on console-image too Nov 09 19:51:18 damn wgere is the rageagainstthecage-arm.bin gone Nov 09 19:54:50 time to attend some talks Nov 09 19:54:58 hem have fun Nov 09 20:15:44 heh.. my plan is broken Nov 09 20:15:56 I can't mount partition from lxc container Nov 09 20:16:04 so I can't newfs it Nov 09 20:17:34 lxc? newfs? Nov 09 20:17:55 my buildbox is lxc container Nov 09 20:38:28 meh Nov 09 20:41:31 hmm Nov 09 20:42:52 conceptually, autoreconf is more about preparation of the source tree than the actual build. were we to share source trees where applicable between recipes, those operations would really need to be done only once Nov 09 20:51:59 GNUtoo|laptop: I saw. And enjoy stuffs were fixed, right pespin\away ? Nov 09 20:52:13 fidencio, they were inproved Nov 09 20:52:26 fidencio, if you want I forward him his mail with the recipes Nov 09 20:52:46 I reviewed his improvements trough irc and critisized the part he left from me which were wrong Nov 09 20:52:48 GNUtoo|laptop: please, send to ML to merge with the upstream Nov 09 20:52:55 fidencio, it's WIP Nov 09 20:53:23 fidencio, if you want I forward you the mail and you're free to do what you want with it Nov 09 20:53:48 GNUtoo|laptop: please. send me. fidencio@profusion.mobi Nov 09 20:53:56 ok Nov 09 20:56:57 fidencio, sent Nov 09 20:57:41 GNUtoo|laptop: thanks! Are you work with pespin? Nov 09 21:21:15 hm.. Nov 09 21:21:37 is there any possibility to prepend bitbake log messages with timestamp? Nov 09 21:21:53 I mean something builtin into bitbake :) Nov 09 21:22:25 using syslog isn't wanted solution.. Nov 09 21:23:31 jay7 normaly the python logging stuff can be configured with timestamp Nov 09 21:26:07 I don't know why it's not an option today Nov 09 21:26:17 It's trivial almost to hack in some form of a stamp Nov 09 21:26:54 yeah should be fairly easy for kergoth Nov 09 21:26:58 to implement it Nov 09 21:27:26 I need this to join data with disk-usage log Nov 09 21:28:56 something like LOG_FORMAT or just PREPEND_TIMESTAMP="some strftime() format" Nov 09 21:29:56 I wish format string because of using unix epoch time :) Nov 09 21:30:45 formatter = logging.Formatter("%(asctime)s - %(name)s - %(levelname)s - %(message)s") Nov 09 21:32:43 Jay7: bitbake master uses the python logging module to make it easier to do that sort of thing Nov 09 21:32:58 for earlier than that, you'd have to monkeypatch bb.msg.note() or something Nov 09 21:33:02 I can't use master because of python 2.5 :( Nov 09 21:33:07 ? Nov 09 21:33:19 ok, I'll look into then Nov 09 21:33:37 or upgrade buildbox to debian-testing Nov 09 21:33:50 you should do that Nov 09 21:34:18 I've used non-stable distribution for a long time to build OE :) Nov 09 21:34:38 I wish some stable period now ;) Nov 09 21:34:51 he pb Nov 09 21:35:00 I run sid for over ten years Nov 09 21:35:54 well.. may be have testbuilder on debian-testing is not so bad idea Nov 09 21:46:31 * Tartarus patches bb.msg.note before Nov 09 21:46:46 it's not bad, for an always on thing anyhow, just need to google up the snippet for a stamp Nov 09 21:48:01 just a time.strftime() should do, just a matter of looking up the format string, i always forget whether its %m for minute and %M for month or vice versa :) Nov 09 21:48:42 well, I'm updating to debian testing and to bitbake master :) Nov 09 21:51:22 Hi all Nov 09 21:52:31 can someone point me what I should change to have timestamps in bb master? :) Nov 09 21:54:31 seems something about logger object Nov 09 21:54:59 mom Nov 09 21:55:38 * Jay7 does known almost nothing of python.. Nov 09 21:55:44 try formatter = logging.Formatter("%(asctime)s - %(name)s - %(levelname)s - %(message)s") Nov 09 21:55:52 make the format as you want Nov 09 21:56:02 okay guys, check this out, i think it should be useful Nov 09 21:56:19 loghandler.setFormatter(formatter) Nov 09 21:56:23 https://github.com/kergoth/openembedded/compare/master...typing Nov 09 21:57:20 you can do.. oe.types.value("PATH", d)[1], assuming you set PATH[type] = "list" and PATH[separator] = ":" Nov 09 21:57:22 hm I thought this was about logging stuff Nov 09 21:58:11 i initially did this just to get config variable validity checking, which that class does, but this could make some things easier in python snippets too Nov 09 21:58:29 for path in oe.types.value("PATH", d): Nov 09 21:58:30 ... Nov 09 21:58:44 thoughts? Nov 09 22:00:11 i was storing the constructed objects in the metadata, but this causes too many headaches unless bitbake itself has more knowledge of it, to handle _append, etc Nov 09 22:00:49 but this is pretty trivial, and i think a step in the right direction. adding the type metadata will make it easier to create configuration UIs Nov 09 22:01:43 woglinde_: should I set handler on every logger=logging.getLogger()? Nov 09 22:02:09 kergoth: is there any single place where I can configure logger? Nov 09 22:03:15 formatting is controlled by the log handler, not the logger. the bitbake UI has a log handler that actually does the display to the console. you could probably import that and modify it from the metadata, or manually edit the knotty.py in bitbake, or you could create your own logging handler to emit your custom format logs to a file rather than modifying the bits going to the console Nov 09 22:04:17 oh.. Nov 09 22:04:18 well Nov 09 22:04:23 let's look into Nov 09 22:04:43 http://docs.python.org/library/logging.html Nov 09 22:05:25 woglinde_: 10x Nov 09 22:07:01 kergoth: is knotty that cmdline UI we using? Nov 09 22:07:27 i.e. when I do "bitbake something" do I use knotty? Nov 09 22:09:29 Jay7: yep Nov 09 22:10:09 well.. then I'll try to add some variables which will control where and what to log Nov 09 22:10:49 may be something alike apache custom log directive :) Nov 09 22:12:23 logging module supports a configuration file Nov 09 22:12:29 just need to add the ability to specify it for bitbake Nov 09 22:16:19 khem, kergoth, i still get "insecure string pickle" after reverting 83cdb23f8b89453a3527a276bd0b4deb85d63deb Nov 09 22:16:53 must be some oe metadata triggering it Nov 09 22:17:16 huh Nov 09 22:19:40 both failure logs i have start in slightly different places, but the context in both appear to be related to ti-biosutils-tree Nov 09 22:19:54 i'll pastebin some pages of context Nov 09 22:20:29 grg: really need to add a temporary catch of that exception in bb.event so we know exactly what event object its failing to pickle. Nov 09 22:20:42 i can make that change, but one of you that can repro would have to run with it :) Nov 09 22:21:26 http://pastebin.com/9xTDu3M7 Nov 09 22:21:36 yeah, i can run it Nov 09 22:45:36 got it! its shr-version Nov 09 22:45:56 bitbake shr-version causes the problem reliably Nov 09 22:46:26 kergoth, ^^^ Nov 09 22:48:23 i can't even -c clean the recipe Nov 09 22:49:08 weird Nov 09 22:49:11 what's the behavior? Nov 09 22:49:26 is this testing-next or master? Nov 09 22:51:36 kergoth, i'm on oe master, but its the same rev as the testing-next tag Nov 09 22:52:30 last i heard, testing-next didn't have the angstrom-version and shr-version fixes for the recursion error Nov 09 22:52:54 http://pastebin.com/Au71WqWV Nov 09 22:53:13 recursion error? Nov 09 22:53:17 grg: is your shr-version exporting METADATA_REVISION? Nov 09 22:53:22 check the recipe Nov 09 22:53:29 yep Nov 09 22:53:39 yeah, that'll explode horribly Nov 09 22:53:43 it shouldn't explode in *this* way Nov 09 22:53:47 but it will explode regardless Nov 09 22:54:23 its due to the fix i put in for the metadata branch and revision, where it uses oe_run instead of os.system(). oe_run emits all exported vars into the env of the subprocess, so if you export METADATA_REVISION, it recurses there Nov 09 22:58:11 indeed a git pull fixes things Nov 09 22:59:20 great Nov 09 22:59:24 still very weird, that error Nov 09 22:59:25 but great Nov 09 23:04:32 03Michael Smith  07master * r30c2729b7e 10openembedded.git/recipes/webkit/ (webkit-gtk_1.2.3.bb webkit-gtk_svn.bb): Nov 09 23:04:32 webkit-gtk: fix packaging of error.html for svn release Nov 09 23:04:32 The resources under ${datadir} have moved around between versions. Nov 09 23:04:32 Use a wildcard in case they move again. Nov 09 23:04:32 No PR bump for v1.2.3. It already packages the correct file. Nov 09 23:04:33 Signed-off-by: Michael Smith Nov 09 23:04:35 03Michael Smith  07master * r5e04a4802c 10openembedded.git/recipes/fbpanel/fbpanel_6.1.bb: Nov 09 23:04:36 fbpanel: add 6.1 Nov 09 23:04:36 Signed-off-by: Michael Smith Nov 09 23:04:37 03Michael Smith  07master * r91eb9425d1 10openembedded.git/recipes/gnome/metacity_2.30.0.bb: Nov 09 23:04:37 metacity: upgrade to 2.30.3 from 2.30.0 Nov 09 23:04:37 Signed-off-by: Michael Smith Nov 09 23:04:44 03Michael Smith  07master * rf27066cf2b 10openembedded.git/recipes/gnome/ (metacity/add-disable-canberra-flag.patch metacity_2.30.3.bb): Nov 09 23:04:44 metacity: patch configure to add --disable-canberra Nov 09 23:04:44 Canberra is still enabled by default, but amenders can add Nov 09 23:04:44 --disable-canberra to EXTRA_OECONF. Nov 09 23:04:45 Signed-off-by: Michael Smith Nov 09 23:08:52 good nite Nov 09 23:09:00 oh.. docbook-utils depend on texlive and lot of x11 libs.. Nov 09 23:09:09 600Mb of crap Nov 09 23:09:39 and do we really needed libsdl1.2-dev? Nov 09 23:09:45 http://wiki.openembedded.net/index.php/OEandYourDistro#Debian Nov 09 23:15:38 okay, here we go. https://github.com/kergoth/openembedded/commit/7ee1902 Nov 09 23:23:54 Jay7: I'd kinda hope not Nov 09 23:23:59 That list needs some cleaning I'm sure Nov 09 23:24:19 I'll try w/o libsdl Nov 09 23:24:37 kergoth: w.r.t. typecheck patch what different types are possible Nov 09 23:24:51 kergoth: I see you use list and boolean ... Nov 09 23:25:00 is it like all python types ? Nov 09 23:25:11 khem: so far: integer, boolean, choice, list, regex Nov 09 23:25:23 they're functions defined in oe.types Nov 09 23:25:24 oh its in code I see Nov 09 23:25:26 yeah Nov 09 23:25:29 yeah :) Nov 09 23:25:34 I was just reading comments Nov 09 23:25:49 i'd have rather unbound them from the metadata, but there are cases when you need to get at flags, e.g. regex supports a regexflags flag, and list has separator, etc Nov 09 23:26:09 ok Nov 09 23:26:12 list is slightly special, as it gives you a subclass that re-joins itself on str() Nov 09 23:26:33 the idea there is that you should be able to str() the object, and feed that back into value() and get an equivalent object back Nov 09 23:26:41 but i'm going to have to fix bool if we want that to work Nov 09 23:26:58 wait, no, that works fine Nov 09 23:27:04 * kergoth forgot that True/False will work as input :) Nov 09 23:27:15 hmmm the basic concept is good imp Nov 09 23:27:17 imo Nov 09 23:32:04 hmm, could change it so the types are independent of the metadata entirely, and store the info about what flags they need as a member of the class, and pass those as keyword arguments to the constructor Nov 09 23:32:18 e.g.: class Foo(object): Nov 09 23:32:23 flags = ["alpha", "beta"] Nov 09 23:32:34 .... Foo(value, *flags) Nov 09 23:32:43 er, **flags Nov 09 23:32:48 * kergoth will think about that Nov 09 23:33:00 would make it nicer to unit tests to reduce the binding to the metadata, i think Nov 09 23:33:03 s/tests/test/ Nov 09 23:38:02 agreed Nov 09 23:38:10 * khem wanders off Nov 10 00:00:17 hm.. lenny was about 700Mb, squeeze - about 1.4G :\ Nov 10 00:07:21 kergoth: Is there any possibility to parse recipes in multiple threads? Nov 10 00:07:28 now it is done on one core :( Nov 10 00:08:05 Jay7: there's a patch for it in my queue from a coworker, has some bugs to be worked out yet Nov 10 00:08:19 cool Nov 10 00:08:26 on my laptop the difference is slim to none, but on this build VM, it drops the parse time from like 4 minutes to like 40 seconds Nov 10 00:08:37 kergoth: I can test it on 6 cores :) Nov 10 00:09:24 kergoth: laptop have an ssd by chance? Nov 10 00:09:33 nah, just slow and only 2 cores :) Nov 10 00:47:48 03Graham Gower  07master * rabd3e10a3e 10openembedded.git/recipes/ (distro-version/distro-version.bb kaeilos/kaeilos-version.bb): (log message trimmed) Nov 10 00:47:48 distro-version, kaeilos-version: don't export METADATA_BRANCH and METADATA_REVISION Nov 10 00:47:48 Causes infinite recursion, which may look like this: Nov 10 00:47:48 NOTE: :maximum recursion depth exceeded while calling a Python object while evaluating: ${@base_get_scm_branch(d)} Nov 10 00:47:49 Or this: Nov 10 00:47:49 Traceback (most recent call last): Nov 10 00:47:50 File "/home/grg/oe2/bitbake/lib/bb/server/none.py", line 128, in idle_commands Nov 10 01:05:46 hmm, anyone know what the performance impact is of using inspect.getargspec() in python? Nov 10 01:05:51 * kergoth 'll have to do some testing Nov 10 01:37:38 say adios to (bb.data.getVar("FOO", d, True) or '').split(), oe.types can handle unset variables == empty list for type == list Nov 10 01:37:39 heheh Nov 10 01:51:51 any idea where this gets an indentation error Nov 10 01:51:53 do_clean_append() { Nov 10 01:51:53 rm -rf ${OECMAKE_BUILDPATH} Nov 10 01:51:53 } Nov 10 01:55:01 do_clean is python, append/prepend aren't that smart, sadly Nov 10 01:55:09 ah Nov 10 01:55:27 some of the other do not look like python Nov 10 01:55:27 try appending oe.path.remove(d.getVar("OECMAKE_BUILDPATH", True)) instead Nov 10 01:56:35 thanks Nov 10 01:56:54 I guess do_clean is one of the tasks that never goes to shell Nov 10 01:57:08 bitbake really needs to be smarter about appending/prepending functions Nov 10 01:57:16 it should assemble a list of functions to execute Nov 10 01:57:20 rather than actually combining their string values Nov 10 01:57:24 do I nee to use tab to indent it? Nov 10 01:57:36 could be, python is strict on indentation Nov 10 01:57:37 :\ Nov 10 01:57:46 much of base.bbclass still uses tabs Nov 10 01:58:02 which also should be fixed.. Nov 10 01:58:42 ok that works Nov 10 01:58:43 thanks **** ENDING LOGGING AT Wed Nov 10 02:59:57 2010